Radial Basis Function Network Outperforms Deep Learning Rivals in Predicting CO2 Emissions Across RCEP Economies

An interpretable artificial-intelligence framework has been developed to predict carbon dioxide emissions across 15 Regional Comprehensive Economic Partnership (RCEP) economies between 2000 and 2024, with a Radial Basis Function (RBF) network emerging as the most accurate and computationally efficient model among fifteen architectures tested. The study, published in Carbon Balance and Management, combines signal decomposition, multi-model benchmarking and explainable AI to turn emissions forecasting into a decision-support tool for carbon management and climate-mitigation strategy.

The researchers note that accurate prediction of CO2 emissions is essential for effective carbon management and evidence-based climate policy, yet emission dynamics are inherently nonlinear and arise from complex interactions among technological, financial, energy, economic and demographic factors.

Fifteen neural-network and deep-learning architectures were then evaluated using mean absolute error, mean squared error, root mean squared error and training time. The Radial Basis Function network achieved the strongest overall performance, with a mean absolute error of 0.028, a mean squared error of 0.002, a root mean squared error of 0.040 and a training time of 00:01:12. The authors conclude that a comparatively parsimonious RBF network can provide accurate and computationally efficient CO2 prediction, an advantage for analysts who need reliable forecasts without heavy computational overhead.

Interpretability was addressed through Shapley Additive Explanations (SHAP) analysis and three-dimensional representation, which revealed heterogeneous contributions from technological innovation, green finance for renewable energy, nonrenewable energy consumption, financial development, economic growth and population aging. Patent activity and research and development expenditure were predominantly associated with lower predicted CO2 emissions, whereas green finance showed context-dependent predictive contributions, meaning its association with predicted emissions varied across settings rather than pointing uniformly in one direction.

By integrating signal decomposition, multi-model benchmarking and explainable AI, the proposed framework is presented as an interpretable decision-support tool for identifying emission patterns and informing targeted carbon-management and climate-mitigation strategies across RCEP economies. The study is an open-access research article published on 7 October 2026, and the version shared is subject to further edits before the final Version of Record. As a modelling study of 15 RCEP economies over 2000–2024, its findings describe predictive associations within that regional and temporal scope rather than causal effects, and the reported associations between factors such as patent activity, R&D expenditure and lower predicted emissions should be read as patterns in the model’s predictions rather than demonstrated policy outcomes.

Subject of Research: An interpretable deep learning framework for predicting CO2 emissions across RCEP economies

Article Title: AI and promote carbon reduction: exploring an optimal deep learning model to predict CO2 emissions under technological innovations and green finance

Article References: Akram, R., Naveed, H. M., Anwer, M. U., Zhang, Z., Song, J., Shahid, M., & Ali, Z. (2026). AI and promote carbon reduction: exploring an optimal deep learning model to predict CO2 emissions under technological innovations and green finance. Carbon Balance and Management. https://doi.org/10.1186/s13021-026-00509-2

Image Credits: AI Generated

DOI: 10.1186/s13021-026-00509-2

Keywords: CO2 emissions, deep learning, Radial Basis Function network, SHAP, green finance, technological innovation, RCEP, signal decomposition, carbon management, explainable AI, neural networks, climate policy